Repository navigation
feat(inference): open Qwen3.5 Agentic coding on FP4+FP8 (same as GLM-5.2/5.3) - #1361
Merged
functionstackx merged 5 commits intoOct 16, 2026
Conversation
…5.2/5.3) Add Qwen3.5 to MODEL_DEFAULT_PRECISIONS so its Agentic coding view opens on FP4 and FP8, matching GLM-5.2/5.3 (SemiAnalysisAI#1215). Both precisions carry full fleets. Explicit i_prec, presets and manual toggles still win; 8K/1K is unchanged. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
@kedarpotdar-nv is attempting to deploy a commit to the SemiAnalysisAI Team on Vercel. A member of the Team first needs to authorize it. |
functionstackx
requested changes
Oct 15, 2026
functionstackx
left a comment
functionstackx
left a comment
Contributor
There was a problem hiding this comment.
thanks for the contribution. sorry on phone and havent had chance to look at it, can u record an before and after video on local dev
Contributor
|
can u paste the videos here in the spirit of open source |
functionstackx
requested changes
Oct 15, 2026
Per review: keep the generic "leaves other models on the auto default" check and point it at a model without an override. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Contributor
Author
|
Before: before.movAfter |
Contributor
Author
|
Test nit addressed in 9039faf: dropped the model-specific test; the generic default check no longer references Qwen3.5. |
functionstackx
requested changes
Oct 15, 2026
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
functionstackx
enabled auto-merge (squash)
October 16, 2026 04:36
functionstackx
approved these changes
Oct 16, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Qwen3.5 397B Agentic coding currently opens on FP8 only, because both precisions clear the 4-curve threshold and the densest-precision rule picks FP8. GLM-5.2/5.3 Agentic already opens on FP4+FP8 via
MODEL_DEFAULT_PRECISIONS(#1215) for the same reason: both precisions carry full fleets.This adds Qwen3.5 to that table. Line labels and point shapes already include precision when both are shown. An explicit
i_prec(including existing FP8-only links), a preset or a manual toggle still wins. 8K/1K and the views API are unchanged.data-mappings.ts: oneMODEL_DEFAULT_PRECISIONSentry for Qwen3.5 Agentic coding.data-mappings.test.ts: covers the new default; Qwen3.5 8K/1K and MiniMax-M3 Agentic stay on the auto default.🤖 Generated with Claude Code
Note
Low Risk
Catalog-only default precision mapping aligned with an existing GLM pattern; no auth or API changes, though removing unit tests reduces regression coverage for this table.
Overview
Qwen3.5 397B Agentic coding now uses the same fixed default precision set as GLM-5.2/5.3: FP4 and FP8 on first load, via a new
MODEL_DEFAULT_PRECISIONSentry indata-mappings.ts. That overrides the densest-precision auto rule (which had been leaving Agentic on FP8 only when both precisions had full fleets).Explicit
i_prec, presets, and manual precision toggles are unchanged. Other Qwen3.5 scenarios and models still use the auto default.The
getModelDefaultPrecisionsunit tests were removed fromdata-mappings.test.ts(including the GLM Agentic cases that lived there).Reviewed by Cursor Bugbot for commit 7dd75a4. Bugbot is set up for automated code reviews on this repo. Configure here.